TWI447842B - Apparatus, device and method for determining alignment errors - Google Patents

Description

Apparatus, device and method for determining alignment error

The present invention relates to a method for determining local alignment error as claimed in claim 1 (due to strain of the first substrate relative to the second substrate when a first substrate is bonded to a second substrate and/or A device caused by distortion. Furthermore, the present invention relates to a device for detecting and/or predicting distortion, particularly for aligning distortions and substrates affecting a substrate, and a method for aligning a first substrate as claimed in claim 4 A contact surface and a second contact surface of a second substrate and/or a device for verifying the alignment, and a method as claimed in claim 8. Furthermore, the invention relates to one of the uses as claimed in claim 9 or 10.

In the earlier European patent application 09012023.9, the basic problems that existed during alignment are described on pages 1 to 3.

Due to the increasingly important 3D technology and the development of miniaturization, it is becoming more and more important to implement a calibration alignment procedure in the combined procedure, especially with a detachable connection prior to the actual bonding procedure. One of the interconnections is called a pre-bonding step coupling. This is primarily important in the application where alignment accuracy better than 2 microns is required for all locations on the wafer. The importance and need for precision in alignment techniques and pre-bonding procedures has increased dramatically to achieve the required accuracy of less than 1 micron (especially less than 0.5 microns or less than 0.25 microns).

Based on the fact that the structure becomes smaller and smaller, but at the same time the wafer becomes larger and larger, there may be a structure that is well aligned with each other near the alignment mark, and at other positions of the wafer, the structures cannot be mutually Properly joined or at least not optimally joined to each other. In order to optimally bond the structures on both sides of the wafer to each other, very complex alignment techniques coupled to a very well monitored and optimized bonding technique, especially pre-bonding techniques, are being developed.

The current technology intends to record some alignment marks on one side of two wafers and then use the alignment marks to align the two wafers and bond the two wafers. Several problems can arise at this time depending on the respective techniques for alignment.

Applicants of European Application No. EP 0 901 202 3.9 have applied a method by which the entire surface of a wafer can be measured to obtain information about the location of the structures on the surface of each wafer.

It is an object of the present invention to develop a general purpose device or a general method that achieves more accurate alignment and minimizes errors in detection position or later alignment, especially with respect to higher detection of the entire surface of the wafer. Furthermore, the object of the present invention is to increase the amount of processing when detecting the position and alignment of the wafer.

This goal was achieved by the features of technical solutions 1, 4 and 8. Advantageous developments of the invention are given in the independent technical solution. All combinations of at least two of the features, technical solutions and/or figures presented in the specification are also included in the framework of the invention. In the range of values given, the values within the indicated limits are also disclosed as boundary values and are claimed in any combination.

The present invention is based on the viewpoint of recording a positional map, a strain map and/or a stress map of the substrates (especially wafers) prior to aligning the substrates, during alignment of the substrates, and/or after aligning the substrates, particularly through the substrates. At least one transparent region of at least one of the plurality of transparent regions to observe them and, as the case may be, in particular, to align the relative positions of the two substrates to each other.

The figure is defined as the property, especially the value, at certain X-Y positions of the substrates along a surface.

The basic idea is to determine at least one strain map along the strain value of at least one of the substrates after joining the substrates, and to determine the local alignment error with the determined strain values. The local alignment error is preferably related to the local structure or a local structural group of the substrates.

As claimed in the present invention, the displacement pattern of the substrate or the two substrates is prepared by the displacement caused by the bonding of the substrates. Such displacements are caused in particular by distortion and/or strain of the substrates.

The basic idea is to be able to estimate how severe the distortion introduced by the pre-bonding step or the bonding step is at a plurality of local locations on the respective substrate, preferably at the location specified by the position map of one of the respective substrates ( →distortion vector). The transparent window can be used to measure the alignment accuracy actually achieved after pre-bonding, but as described below, the distortion can be degraded due to such distortion, thus indicating the actual accuracy achieved over the entire wafer. Only a small part. Since the wafers are opaque to infrared radiation, alignment accuracy cannot be directly measured. According to the invention, this is estimated by detecting a stress map and/or a strain map.

An important aspect is that in a preferred embodiment, the device (or measuring device) as claimed in the present invention is provided as a separate module from the alignment device.

In a preferred embodiment, the module is divided as follows:

1) Module for detecting a stress map and/or strain map before joining (bonding or pre-bonding)

2) Alignment modules, in particular according to European Patent Application EP 0 901 202 3.9. However, wafer alignment can also occur using only two alignment marks. In this case, the location map will be detected by real measurement, but can also be known based on the wafer layout.

3) At least one measurement module for detecting a stress map after bonding.

In addition to the present application, as embodied in the present invention, another embodiment, also described hereinafter in the present invention, is also conceivable, wherein one of the two wafers is not heavily structured (ie, has the largest pair) Quasi-marking). In this case, it is possible to estimate the distortion of the structured wafer. In this example, there is no "measured" location map, but only information or information about the existing distortion of the exposure field and information about the exposure fields on the wafer. These materials are "read" and are known from the wafer layout (location) on the one hand. These existing distortions are measured with a measuring device suitable for this purpose (for this purpose, a lithography system is typically used).

In this embodiment, there may be less focus on alignment (accurate alignment) (one of the wafers is not heavily structured), but only the distortion of the structured wafer is related. The alignment between the two wafers is here only for imprecise alignment (mechanical-edge to edge) or for optical alignment (via alignment marks applied to wafers that are not heavily structured).

However, the need for optical alignment is typically quite low.

The location maps are recorded/detected in an advantageous version, as described in the earlier application EP 09012023.9.

The prior application EP 0 901 202 3.9 describes a method in which the XY position of the alignment keys of the two substrates to be aligned can be detected or measured in at least one XY coordinate system independent of the movement of the substrates, such that The correlation of the associated alignment keys of a second substrate aligns the alignment keys of a first substrate to corresponding alignment positions. Therefore, a position map of one of the substrates to be aligned is prepared.

In other words: in particular in the X and Y directions, the device uses a member for detecting the movement of the substrate, the members designating at least one fixed (especially partially fixed) reference point and thus at least in the X and Y directions The upper one enables accurate alignment of the corresponding alignment keys.

The location map can be recorded as follows:

Configuring a first contact surface in a first X-Y plane and a second contact surface in a second X-Y plane parallel to one of the first planes X-Y,

- detecting, by the first detecting component, the XY position of the first alignment key positioned along the first contact surface in the first XY coordinate system independent of the movement of the first substrate and by the second detection The member detects an XY position of the second alignment key positioned along the second contact surface and corresponding to the first alignment key in the second XY coordinate system independent of movement of the second substrate,

Aligning a first contact surface in one of the first alignment positions based on the first XY position determinations and aligning with respect to the first contact surface and determining a second pair based on the second XY position determination a second contact surface in the quasi-position.

Rather, if nothing is described herein, this is also particularly applicable to wafer and coordinate systems on recording and moving platforms and to the relationship between the strain patterns and/or stress maps that can be used to record/detect.

By combining a position map, a strain map and/or a stress map (especially in combination with a transparent region), especially after the pre-bonding step or during the pre-bonding step, it can be detected after contact with the wafer or due to contact with the wafer The defects are aligned and the wafers can be separated from each other again or can be separated from the production process.

One problem is that by the prior art, only a very limited number of alignment marks is detected by convention. Conventionally, alignment is performed using only two alignment marks. This may then result in the inverse effect described above: the alignment at the location of the alignment marks and the location in the region directly adjacent to the alignment marks may be good, while the remaining area of the wafer The alignment in may be inappropriate.

Another problem is that mechanical distortion can occur on one or two wafers depending on the selected bonding procedure of the pre-bonding of the wafers and the final combination of the wafers, which may be local or even Overall degraded alignment accuracy. The importance/effect of such distortion relative to the successful alignment of the wafers increases with the required alignment accuracy, particularly to achieve the desired accuracy better than 2 microns. For alignment accuracy greater than 2 microns, these distortions are not small enough to represent a significant effect on one of the alignment results.

Such distortions not only create a problem when combining two structured substrates, but can also cause major problems when bonding a structured substrate to a substrate that is not heavily structured. After bonding, it will be particularly advantageous to implement other program steps that require a very precise alignment of the structured substrate. In particular, the lithography technique in which the additional layers of the structure are aligned with the existing structures of the structure should now impose high demands. These needs increase with decreasing the structural size of the structure being produced. This application is produced, for example, in the production of so-called "back lighting CMOS image sensors." At this point, one of the first wafers having a structured surface is bonded to one of the carrier wafers, particularly not heavily structured. After forming a permanent bond connection, removing a majority of the wafer material of the structured wafer allows the structured surface (especially the light sensitive sites) to become accessible from the back. This surface must then be subjected to other program steps (especially lithography) to, for example, apply the color filters required to operate the image sensor.

Distortion of such structures adversely affects the alignment accuracy that can be achieved in this lithography step. For the current generation of image sensors with a pixel size of, for example, 1.75 microns or 1.1 microns, one of the first exposure fields (up to 26 x 32 mm) and the repeated exposure system allow for a distortion of roughly 100 mm, better It is 70 nm or 50 nm.

The pre-bonding in this document specifies a bond connection that also allows separation of the substrate (especially the wafer) after a complete pre-bonding step without causing irreparable damage to the surface. Therefore, such bonding connections may also be referred to as reversible bonding. This separation is possible based on the fact that the bond strength/adhesion between the surfaces is still relatively low. This separation is possible until the binding is permanent (i.e., no longer separated (non-reversible)). This is achieved in particular by the effect of externally on the wafer over a certain time interval and/or via physical parameters and/or energy. In this case, it is suitable to compress the wafers, for example, via a compressive force or to heat the wafers to a certain temperature or expose the wafers to microwave radiation. An example of such a pre-bond is a connection between a wafer surface having one of the thermally generated oxides and a surface of the wafer having one of the native oxides, and van der Waals between the surfaces caused at room temperature at this time. (van-der-Waals) connection. These bond connections can be converted to permanent bond connections by temperature processing. Advantageously, such pre-bonded connections also permit inspection of the binding results prior to forming the permanent bond connection. In the case of the deficiencies identified in this test, the substrates can be separated again and rejoined.

In the simplest embodiment, the present invention is a measurement device and a measurement method that enables detection of stress introduced by a pre-bonding step in a wafer or wafer pair. This occurs by analyzing the stress maps before and after the combination. A stress difference map is generated therefrom according to the following description.

This stress difference map enables the estimation of the distortion/strain introduced by the pre-bonding step, especially empirically. From it produces a distorted vector field or a distortion map / strain map.

As claimed in the present invention, this distortion vector field enables wafer pairs in which one of the two wafers is structured to determine that certain locations are generated, in addition to the deviation from the ideal shape that existed prior to bonding. The distortion, especially at the center of the exposure field, is preferably at the location of the alignment marks for the lithographic device.

Alternatively, the distortion vector field causes wafer pair prediction with two structured wafers to be theoretically desired (based on the position map of the two wafers as a result of the selected ideal alignment position) The deviation vector is the extra alignment deviation expected at the point detected in the position maps. This produces a bias vector field or a displacement map.

In a preferred embodiment, the desired deviation vector field is superimposed or added to the deviation that has been determined based on the measurements in the transparent window. This, in turn, results in a final desired alignment result for all of the corresponding locations provided by the location map. With this result, a decision can be made as to whether or not to separate the bonded wafers again.

Moreover, the present invention is based on the idea of designing a device and a method, wherein:

- The alignment of the two wafers with respect to one another can be determined, wherein all of the structures on the contact surfaces of the wafers are economically or technically optimal for each other. This relative position may (but is not necessarily) correlated with one another with excellent alignment of one of the alignment marks. Of course, most of the alignment marks are also always in the most suitable position (i.e., at least relative to the immediate micrometer range, but not necessarily optimal).

- For the completed "pre-bonding" procedure, therefore, in a state in which the two wafers can also be separated from each other, the stress generated in the pre-bonding step and the distortion that may originate therefrom (especially mechanical distortion) can be inspected Is it at an acceptable level? This is especially useful in applications where one of the two wafers is structured and the second wafer is not heavily structured.

- for a completed "pre-bonding" procedure, therefore, in a state in which the two wafers can also be separated from each other, the accuracy of positioning the two wafers or the individual structures of the wafers to each other can be checked Whether the accuracy actually corresponds to the specification. In this way, it is possible to determine the displacement of the wafers as they approach each other in the z-axis direction or (or even worse) the deviation from the ideal position that occurs during the contact procedure. In particular, due to the detection of the stress introduced in the pre-bonding step, the prediction of the desired distortion and the resulting deviation from the ideal position can also be estimated, inter alia, empirically.

With this device and this method, alignment accuracy of less than 25 microns (especially less than 0.15 microns, preferably less than 0.1 microns) can be accomplished by controlling and correcting the distortion described above prior to the creation of the final bond connection. With good reliability and yield.

In other words, therefore, the device uses at least components for detecting the stress properties of the wafer before and/or after the pre-bonding step. Based on the knowledge of these stress properties and one of these stress properties, especially before and after the pre-bonding step, predictions can be made regarding strain/distortion that has been introduced into the wafer during pre-bonding.

In another embodiment, particularly for inspection and/or alignment of two structured wafers, the device uses components for detecting movement of the substrate, particularly in a single X and Y direction, the components specifying at least A fixed (especially locally fixed) reference point and thus enables accurate alignment of corresponding alignment keys in at least one of the X and Y directions, not only with reference to the position of the respective substrates, but also to the strain and/or stress properties.

The features of the invention presented herein are in one aspect in that it is possible to determine, measure and/or inspect all of the structures of the two wafers for economical and/or technically optimal alignment with each other. This includes: recording a positional map of the structure of the two wafers prior to concentrating the wafers in accordance with earlier European Patent Application No. 09012023.9; and continuously monitoring, in particular, on-site monitoring of the displacement of the two wafers via alignment marks The program. For defective pre-alignment and pre-bonding, generally expensive structural wafers can be separated and realigned again from each other.

In an advantageous embodiment of the invention, it is provided that the device is manufactured to take into account the first position map of the first alignment key and/or the second position of the second alignment key by means of the evaluation means, in particular at the time of determination Figure. The alignment keys are especially alignment marks and/or structures applied to the substrate or one of the substrates.

The stress maps and/or strain maps are recorded from the respective inspection sides by reflection measurements, and the radiation is reflected. In particular, one of the stress/strain averages over the thickness of the layer is not available, but as suggested by the present invention, information about the area near the surface, infrared light, or from the respective back by transmission measurements can be obtained. In the measurement using infrared light or x-ray, an average value of stress or strain throughout the thickness of the detectable layer is determined. A stress-strain diagram is recorded without passing through the transparent area. In particular, the position map is determined solely by reflection measurement, preferably by using visible light. The first alignment keys and the second alignment keys can be simultaneously detected by the same detection component through the above measurement.

According to a further advantageous embodiment of the invention, it is provided that, in particular, the alignment of the substrates during contact and/or bonding of the substrates can be inspected in the field. The advantage of this on-site inspection is that the alignment errors caused by moving the substrate during contact or bonding can be ruled out.

To a certain extent, four corresponding alignment keys are provided for inspection, especially through transparent areas, and the relative positions of the substrates to each other can be simultaneously monitored on site.

Other advantages, features, and details of the present invention will become apparent from the following description of the preferred embodiments.

In the figures, the same components and components that have the same function are identified by the same reference numerals.

Figure 1a shows a typical wafer system consisting of a first substrate 1 (especially a wafer 1) having a surface 1o and a second substrate 2 (especially a wafer) having a surface 2o. On the surfaces 1o, 2o are different structures 50, 50' bonded to the contact surfaces 1k, 2k. The structures 50, 50' can be, for example, holes, among which are MEMS devices. In the case of a 3D integrated wafer stack, the structures may also be metal surfaces for making electrical connections. For the sake of simplicity, the structures 50, 50' are shown as black rectangles. Figures 1b and 1c show the surfaces 1o, 2o of the two wafers 1, 2. The wafer 2 has four regions 40 with second alignment keys 4.1 to 4.4.

These regions 40 are transparent to electromagnetic radiation of a certain wavelength or a certain wavelength range. A first detecting member 7 (particularly an optical device) can relate the first alignment keys 3.1 to 3.4 of the first substrate 1 to the corresponding second alignment keys 4.1 to 4.4 through the transparent regions 40. Advantageously, such transparent regions can be maintained at relatively low values by the extent to which doping or especially doping is avoided for such regions and that no metal layer is used in such regions or in particular produces relatively few metal structures. Useful for wafers. This can be achieved, for example, in that only the alignment marks and possibly related structures, which may be composed of metal, are placed in the transparent regions. Subject to these prerequisites, 矽 is transparent to infrared light having a wavelength greater than 1 micron (especially greater than 1050 nm).

The structures 50, 50' may protrude from the surfaces 1o, 2o or may be reset relative to them, for which reason the contact surfaces 1k, 2k need not be 1o with the surfaces of the wafers 1, 2 2o coincide.

The alignment keys 3.1 to 3.n or 4.1 to 4.n may also be part of the structures 50, 50' or the structures 50, 50'.

The method begins by recording a location map. A position map is defined as (as spatially as complete as possible) position detection of as many structural elements as possible, in particular the first alignment keys 3.1 to 3.n on the surface of the wafers 1, 2 And/or position detection of the second alignment keys 4.1 to 4.n and/or the structures 50, 50' or portions of the structures 50, 50'.

Figure 2a shows the position detection of the surface 1o of the first wafer 1 by the optical device 7, thus recording a first position map. The position of the first alignment keys 3.1 to 3.4 is measured on the top 1o of the wafer 1 by moving the wafer 1 relative to the optical device 7 or moving the optical device 7 relative to the wafer 1. In a preferred embodiment, the optical device 7 is attached to the recording member 12 to move the wafer 1 relative to the optical device 7.

In a second step, in particular after the first step or simultaneously with the first step, according to FIG. 2b, the top 2o of the second wafer 2 is implemented via a second detecting member, in particular optics 8. The same procedure.

Since the recording position map is important in this measurement procedure, it is also conceivable to use only the optical device 7 as the detecting member, so that the optical device 8 is omitted and the top portion is structured in the direction of the optical device 7 The two wafers 1, 2 are measured by 2o. For a later alignment and bonding step, one of the two wafers 1, 2 is then flipped and secured to the recording member 12 or 22.

According to the steps described above, the device now knows all recorded structures 50, 50' on the tops 1o, 2o of the wafers 1, 2 or the recorded first alignment keys 3.1 to 3.n. And the XY positions of the second alignment keys 4.1 to 4.n, and it is also known that the structures 50, 50' are relative to the first alignment keys 3.1 to 3.n and the second alignment keys 4.1 to 4.n location. It is stored in the form of a first position map of the first substrate 1 and a second position map of the second substrate 2.

As claimed in the present invention, during the measurement step, not only the first position map and the second position map but also the first and second initial strain maps in a different module or simultaneously in a module are recorded. / or first and second stress maps, and they represent the basic stress or initial stress of the substrates 1, 2. At this time, the strain value and/or the stress value are recorded as a function of the X-Y position according to the position map. Various measuring devices, in particular infrared measuring devices, capable of determining locally resolved stresses and/or strains can be used. It is especially advantageous to use a measurement device based on Raman spectroscopy. Alternatively, as claimed in the present invention, an infrared method "Grey-Field Polariscope" "Review of Scientific Instruments 76, 045108 (2005)" "Infrared grey-field polariscope: A tool for rapid stress analysis in microelectronic materials and devices" can be used. . The stresses and/or strain maps are then recorded by the relative motion of the wafers 1, 2 by the optical devices 7, 8. In an advantageous embodiment, there are separate optics or optics additionally integrated into the optics 7, 8.

To a certain extent, the preparation of only the strain map or the stress map stress map for the optimization of the detection time can be converted into the corresponding strain map via the basic equation of the elastic principle, and vice versa. As claimed in the present invention, it is conceivable to have a mathematical (especially numerical) conversion of one of the starting points of the method according to the finite element.

For devices that have been optimized for particularly accurate detection of position maps and/or strain maps, two different detection components are used to detect position maps and/or strain maps.

In accordance with the present invention and to exclude other sources of defects, it is provided that the stress map and/or the strain map are detected based on the alignment of the substrates 1, 2.

The respective detecting means for simultaneously recording the position map in an advantageous configuration includes detecting means for detecting the stress map and/or the strain map such that the respective detecting members having the same drive move.

Alternatively, in order to speed up and in this regard more cost-effective embodiments, it is conceivable to provide detection of stress maps and/or strain maps in one or more separate modules, in particular by separate wafer handling components (compare Good for the robotic arm).

Figure 3 shows the first and second alignment keys 3.1 to 4.1, which are excellently aligned, and the highly aligned structures 50, 50' which are covered by the structure 50 due to excellent overlap. . Unrealistically, the state from Figure 3 results in the excellent production of two wafers 1, 2 with reference to the alignment keys 3.1 to 3.n and 4.1 to 4.n in an excellent combination procedure. All structures 50, 50' above. In fact, the structures 50, 50' cannot be produced with such accuracy. Even if the structures 50, 50' are so excellently produced, the wafers 1, 2 can move relative to each other during the proximity procedure or when the contact is initiated. In the pre-bonding step, additional strain can also be introduced into the wafer, which causes strain/distortion and is a further consequence of one of the ideal alignment deviations. As claimed in the present invention, excellent alignment at individual locations is therefore not necessarily an end goal. Specifically, as claimed in the present invention, care should be taken that all corresponding structures 50, 50' on wafers 1, 2 are fully aligned with respect to economic and/or technical aspects such that for each wafer to be bonded and aligned Yes, the grain waste is as small as possible.

Since the locations of all of the detected structures 50, 50' and/or the first alignment keys 3.1 to 3.4 and the second alignment keys 4.1 to 4.4 of the wafers 1, 2 are known, The optimum relative position of the wafers 1, 2 to each other or the optimum relative position of all of the structures 50, 50' to each other is determined by the computing means. This occurs by determining a first alignment position of the first contact surface 1k and a second alignment position of the second contact surface 2k based on the value of the first position map and based on the value of the second position map. During the contact and also after the contact and during the bonding process and after the bonding process, the relative position of the wafers 1, 2 and each other can be continuously inspected in the field by the optical device 7 and through the transparent region 40 for correctness. / or the first alignment position and the second alignment position. In this way, alignment can be checked on site.

For example, the optimum relative position of the two wafers 1, 2 to each other or the structures 50, 50' is generated by calculating a minimum sum of one of the respective quadratic deviations of the respective corresponding structures 50, 50' The most suitable relative position for each other.

As claimed in the present invention, it is also conceivable that the economical aspect is also included in this calculation of the ideal alignment position. Therefore, in many areas of the semiconductor industry, especially in the memory industry (eg, RAM, NAND flash memory), wafers on certain areas within the wafer (especially in the center of the wafer) have The number of variations in small quality related parameters. Therefore, wafers originating in this area can achieve higher selling prices such that the classification procedure (this procedure is referred to as "grading") is considered, wherein such wafers are intentionally divided into different quality baskets. Advantageously, therefore, as claimed in the present invention, the ideal alignment position of the wafers is not only calculated based on the position maps of the two wafers, but also an economic calculation/configuration, which should be particularly cautious: A higher yield in the region of the higher quality wafer is achieved at a lower yield in the region of the lower value wafer.

4 and 6 show a difference vector u that differs between the X-Y positions of the corners of an upper structure 50 and the corresponding lower structures 50'. For example, the difference vector u is generated from the minimization calculation of the position map. Of course, in each of the transparent regions 40, its own difference vector u can be identified. At this point, if the two wafers are close to each other, the difference vector u is continuously checked. If the difference vector u changes during the approach or during the contact or bonding, the deviation of the determined relative positions of the two wafers 1, 2 from each other occurs. Even if the two wafers 1, 2 are in contact, the optical device 7 can inspect at least four difference vectors u through the transparent regions 40. If one of the deviations is too large after the contact, the wafers 1, 2 are immediately separated to perform the alignment and pre-bonding procedure again. To perform simultaneous inspection of the plurality of transparent regions 40, for each of the transparent regions 40, there is its own optics 7 such that the amount of processing during bonding is not reduced by on-site detection of alignment.

Alternatively, it is also conceivable as claimed in the present invention that the inspection step is performed after pre-bonding of the separation modules such that the throughput of the alignment members and the throughput of the pre-bonded modules are not reduced. The possible separation of the wafer after the inspection step can occur in the module to be inspected, but can also occur in a separate module. It is also conceivable that not all of the modules are connected in a single device, but in particular the separation means is formed by the wafer processing members thus separated.

Figure 3 shows an enlarged view of one of the close regions of alignment marks 3.1 and 4.1. To be able to detect structures 50 and 50' during stacking, only the edges are shown. If at this point the structures 50 and 50' are perfectly oriented with respect to each other and the alignment marks 3.1 and 4.1, an excellent coverage of one of the two structures 50 and 50' will be established in a bonding procedure.

Figure 4 shows the same situation in which the structures 50 and 50' do not have coverage, but the alignment marks 3.1 and 4.1 have been perfectly aligned with each other. In the enlarged view of the structure 50 and the corresponding structure 50', it is recognized that the difference vector u has one X component and one Y component usable for vector calculation.

An important component of the present invention is that the measuring instrument provided in the above-mentioned measuring instrument or a separating module can be used for stress measurement and/or strain measurement after pre-bonding or combining, The stress and/or strain of the bonded wafer stack is determined. Measuring the initial stress pattern and/or initial strain map of the wafers 1, 2 before the wafer stack by measuring the two wafers 1, 2 and measuring the stress map and/or strain of the wafer stack The graph can be concluded with respect to the deformation at the moment of contact or a conclusion can be drawn shortly. In other words, the stress introduced by the pre-bonding procedure can therefore be measured and the stress/distortion can be determined/estimated/predicted or advantageously calculated, inter alia, based on empirically determined relationships.

Although the inner regions of the wafers 1, 2 are no longer viewed by the optical devices 7, 8, but since there is no transparent region in this region, it can also be obtained from the region by strain patterns and/or stress maps. The conclusion of the state, position or deformation. If, for example, a stress prevails over a critical value (e.g., a value of a comparative stress) in a region, the region can be automatically marked by the software as a problem region. Therefore, the crystal grains can be classified into quality categories. Grains with low intrinsic stress have a good quality class and long service life, while grains with high stress concentrations can be classified into a low quality category.

Based on these stress maps/strain diagrams, the alignment accuracy that has been achieved is estimated and empirically determined for the entire wafer surface and all structures present thereon. This can be done in practice as follows.

1) Detecting a first position map and a second position map corresponding to the first wafer and the second wafer as described above.

2) Calculate the ideal alignment position based on the first position map and the second position map according to technical and/or economic criteria. This calculation also produces a desired alignment position and a corresponding deviation vector for the alignment marks in the transparent region 40. That is, it is not necessarily optimal to align the alignment marks in the transparent regions 40 to achieve the most appropriate result for the entire wafer. Furthermore, based on this calculated desired alignment position, a two-dimensional difference of the individual difference vectors of at least the main number (preferably all positions included in the position map) expected for this reason (see Fig. 6) is calculated. Vector field v'. Here, preferably, the sites without the structures 50, 50' are omitted to not confuse the measurement results. For example, due to the alignment marks (rather than the structures 50, 50'), the alignment positions are the positions of the alignment keys 3.1 to 3.4 and 4.1 to 4.4.

3) Detecting a first initial stress map and a second initial stress map corresponding to the first wafer and the second wafer before the pre-bonding step, and particularly detecting the first position map and the second position map in parallel.

4) Pre-bond the wafer in a suitable manner. For most variations of bonding, these methods are known in the basic form to those skilled in the art.

5) Detecting the actual alignment accuracy in the transparent region 40 and determining the actual deviation vector u in the transparent region 40.

6) Determine the difference between the actual deviation vector u for the transparent region and the calculated deviation vector.

7) Considering the determined difference, the resulting difference vector map v" can be calculated, wherein for at least the main number, especially for each position included in the first position map and the second position map, there is a deviation vector corresponding to the individual position. The deviation vectors u are now adapted by calculating one of the correction vectors for the respective individual positions based on the deviation vectors determined at step 6 and the coordinate positions of the respective points and the coordinate positions of the transparent fields.

8) Detecting the first stress map and the second stress map after pre-bonding.

9) Compare the first stress map before and after pre-bonding with the second stress map before and after pre-bonding.

10) Predict the additional alignment error/deviation vector desired for individual points based on stress differences before and after pre-binding.

11) Add the additional deviation vector caused by the stress introduced in the pre-bonding to the deviation vector theoretically expected and calculated in step 7.

12) determining whether the alignment accuracy is desired (here, based on the calculation in step 11 for the deviation vector in the individual points as described above corresponding to the technical and economic success criteria), or determining whether to implement the wafer re Processing / separation.

For wafer stacking, where one or both wafers prior to pre-bonding have only low initial stress or especially no initial stress or initial stress prior to known bonding, because it is only experienced, for example, in mass production Very low variation, in step 3, the detection of the stress map prior to bonding can be omitted for the purpose of optimizing throughput and cost. As claimed in the present invention, it is also particularly possible that the stress undergoes only a low change to only partially detect the stress pattern before the bonding. In this connection, low stress is defined as a stress value that is negligible compared to the stress generated in the pre-bonding step. This is especially the case when the stress difference is 3 times, the preferred difference is 5 times, or even better, the difference is 10 times. Relative to the partial measurement of the wafer stack, for example, subjecting a batch of the first wafer stack and the last wafer stack to inspection and using the stress map determined by the first wafer stack for the remaining wafer stack It is especially feasible for calculations. It is also conceivable to calculate the offset in real time to then calculate an average stress map for the first wafer stack and the last wafer stack based on the calculation, for example. In this case, it may also be advantageous to additionally inspect other wafer stacks to achieve higher reliability in calculating the average. According to the described procedure, only one of the two wafers forming the wafer stack can be subjected to stress pattern detection. This is particularly advantageous when only one of the two wafers does not meet the criteria described above, which proves that omitting the stress map detection is reasonable.

For applications that structure only one of the two wafers, the method can be performed similar to combining two structured wafers. Specifically, the program in this embodiment is as follows:

1) Detecting existing distortion/deviation vectors for individual exposure fields of the ideal shape on the structured wafer by means of suitable detection means. In particular, it is also intended that the subsequent processing of the bonded wafer and the repeated exposure system with the aid of a suitable device, such as a test mark, enables such measurement. This deviation from the ideal shape is represented in the form of a vector field and stored for further calculation. In particular, the vector field contains a vector of major portions, including in particular all locations of the alignment marks that are conventionally located at the corners of the exposed fields.

2) Before detecting the pre-binding procedure by a suitable detection component from the side opposite the contact surface 1k (if the wafer 1 is a structured wafer) or the contact surface 2k (if the wafer 2 is a structured wafer) The initial stress map of the structured wafer.

3) Aligning the two wafers with each other with the aid of suitable sensing members for the wafer position and alignment members.

4) Pre-bond the two wafers.

5) Detecting the stress map of the structured wafer after the pre-bonding step via a suitable detection member from the side opposite the contact surface 1k/2k.

6) Determine the difference between the stress map before the pre-bonding step and the stress map after the pre-bonding step.

7) Deriving the desired distortion vector/desired distortion vector field based on the difference in stress determined in step 6. Advantageously, the vectors in this vector field are determined for the position associated with the position of the vector from the vector field that has been determined in step 1, in particular at least in large scale. Advantageously, this configuration is better than 500 microns, but more desirably better than 200 microns or 700 microns.

8) Add the distortion vector field and the vector field determined in step 1.

9) Check if the vector field calculated from step 8 corresponds to the technical and economic success criteria as described above or whether the separation and reprocessing of the wafers occurs.

The partial detection of the stress maps previously described or omitted for the selected wafer and/or wafer stack prior to the combination of the stress maps prior to bonding is approximated at this time.

As claimed in the present invention, a distortion vector field can be derived from a stress map based on a plurality of suitable methods and a stress difference map between pre-bonding and post-bonding can occur in particular. The detection of the self-stress map and, in particular, the difference in stress before/after some areas of the wafer are apparent, and an additional stress or tensile stress has been generated during the bonding. On this basis, conclusions can be drawn about the direction of the individual vectors at any point of the wafer. The same level of self-measurement and/or calculation of the stress difference in the known individual regions allows conclusions to be drawn about the amount of vector. However, since the individual component regions of the wafer are conventionally surrounded by other regions that additionally affect the strain/distortion of the wafer, such relationships are not necessarily linear. Therefore, a complete computing module suitable for practice must be used to be able to estimate the actual amount and direction of the vectors. Another possibility for certain profile conditions (some stress values, etc.) also uses empirical methods, which utilize findings from tests that have been completed in the past.

As shown in Figures 5a through 5c, in the absence of the transparent region 40, the on-site measurement of the alignment during contact and/or bonding is limited to the measurement of the strain field and/or the stress field. The example of Figures 5a through 5c shows a method and a device in which a structured wafer 2' is aligned with respect to a carrier wafer 1', rather than two wafers 1, 2 that are both fully structured.

The alignment marks 3.1 to 3.n are associated with the alignment marks 4.1 to 4.n by using a known optical system. If the optical devices 7 and/or 8 have the corresponding sensor members mentioned above, they can be used to measure the strain field and/or the stress field. If the electromagnetic radiation used can penetrate the structural wafer 2' and the carrier wafer 1' has been removed from the visual area (Fig. 5b), the optical device 8 or the optical device 7 can measure the Stress field and/or strain field at the top 2o of wafer 2'. By calculating the component it must be considered that if the stress is varied along the layer thickness (so-called stress gradient in the layer thickness) by the transmission measurement of the optical device 7, an average strain and/or stress value can be obtained. Referring to Figure 5c, the necessary changes to the application described above have been made to measure the strain field and/or stress field on the surface 1o of the carrier wafer 1'.

After measuring the respective initial strain fields and/or stress fields, the two wafers 1', 2' can be aligned and bonded. The strain fields and/or stress fields are determined via the optics 7 and/or 8 after the bonding is completed. After bonding, since electromagnetic radiation must penetrate the two wafers 1', 2', only one transmission measurement of the strain fields and/or stress fields of the surfaces 1o, 2o is possible. Therefore, the aforementioned difference between the transmission measurement and the reflection measurement is preferable. As suggested by the present invention, for better comparison, transmission measurements are preferred. If the transmission measurement and the reflection measurement should produce similar strain maps and/or stress maps, it can be inferred that the strain fields and/or stress fields are only on the surfaces 1o, 2o and have no stress gradient throughout the thickness. The previous/after comparison then allows conclusions to be drawn about one of the changes in the strain fields and/or stress fields and one of the possible shortcomings of the system. If the ultimate strain region and/or stress region or the like is found to exceed a comparison value, the wafer system can be decomposed into individual wafers again before they are permanently bonded to each other.

For structured wafers that are opaque to the electromagnetic waves used to detect the stress map, a reflectance measurement may be preferred because the transparency of the structured surface of the wafer (especially the contact surface 1k or 2k) does not function. For such wafers having transparency, stress can also be advantageously measured on the surface opposite the surfaces 1o and 2o. To achieve a better comparison of the measurements, it is a good idea to measure these surfaces before and after the pre-bonding and/or combining steps. Since the stress field in the wafer viewed in the lateral direction has a larger spread than the wafer thickness, this version of the measurement also produces very good results. In particular, an environment with a certain minimum extended lateral stress to cause significant distortion is desirable for accuracy. It may be desirable to have a stress field in the lateral direction (X/Y) that is at least 3 to 5 times, and possibly even 10, 15 or 20 times, relative to the wafer thickness extension, to cause associated strain/distortion.

As claimed in the present invention, the most important combination of wafer materials that can be used is: Cu-Cu, Au-Au, hybrid bonding, Si, Ge, InP, InAs, GaAs; and combinations and materials that allow such materials Each of them can be assigned an oxide.

Advantageously, the position map, strain map and stress map are all about the same X-Y coordinate system. Therefore, the vector calculation is simplified, especially when the first alignment position and the second alignment position are determined according to FIG. 6 and the displacement map is determined.

1. . . First substrate

1'. . . Carrier wafer

1k. . . First contact surface

1o. . . Surface/top

2. . . Second substrate

2'. . . Structured wafer

2k. . . Second contact surface

2o. . . Surface/top

3.1 to 3.n. . . First alignment key

4.1 to 4.n. . . Second alignment key

7. . . First detecting member

8. . . Second detecting member

40. . . Transparent area

50, 50'. . . structure

u. . . Difference vector

v', v"... deviation vector

Figure 1a shows a cross-sectional view of one of the aligned wafer pairs as claimed in the present invention;

Figure 1b shows a schematic representation of one of the upper wafers according to the wafer pair of Figure 1a;

Figure 1c shows a schematic representation of one of the lower wafers of the wafer pair according to Figure 1a;

2a is a schematic diagram showing a procedure of detecting a first wafer as claimed in the present invention;

2b is a schematic diagram showing the steps of detecting a second wafer as claimed in the present invention;

2c is a schematic view showing the alignment of the wafer in the field as claimed in the present invention when the wafer is in contact;

Figure 3 shows an enlarged view of one of the alignment marks for a wafer pair that is excellently aligned and in contact;

4 shows an enlarged view of one of the centers of two structures of an alignment of one of the aligned pairs of wafer pairs aligned and in contact with each other;

5a-5c illustrate an alternative method for detecting alignment of a wafer pair; and

Figure 6 shows a schematic diagram of one of the determinations of a displacement map as claimed in the present invention.

1. . . First substrate/wafer

1k. . . Contact surface

1o. . . surface

2. . . Second substrate/wafer

2k. . . Contact surface

2o. . . surface

3.1. . . First alignment key

3.3. . . First alignment key

4.1. . . Second alignment key

4.3. . . Second alignment key

7. . . First detecting member

40. . . Transparent area

50. . . structure

50'. . . structure

Claims (14)

A device for determining a local alignment error due to the first substrate (1) being opposite to the second when a first substrate (1) is bonded to a second substrate (2) Caused by strain and/or distortion of the substrate (2), which has the following feature: detecting the strain value along the first contact surface (1k) of one of the first substrates (1) by the initial detecting member a strain map and/or a first stress map along a stress value of the first contact surface (1k); and/or detecting a strain value along a second contact surface (2k) by the detecting member a second strain map and/or a second stress map along one of the stress values of the second contact surface (2k); and may be used to evaluate the first strain map and/or the second strain map and And/or the first stress map and/or the evaluation component of the second stress map determine the local alignment errors. A device as claimed in claim 1, which is manufactured to take into account a first position map and/or a second alignment key (4.1 to 4) of the first alignment key (3.1 to 3.n) by means of the evaluation member, in particular at the time of determination .n) The second position map. A device for determining a local alignment error due to the first substrate (1) being opposite to the second when a first substrate (1) is bonded to a second substrate (2) Caused by strain and/or distortion of the substrate (2), which has the following feature: one of the strain values along the first contact surface (1k) of the first substrate (1) can be detected by the initial detecting member a first initial strain map and/or a first initial stress map along one of the stress values of the first contact surface (1k); and/or detectable along a second contact surface by the initial detecting member (2k a second initial strain map and/or a second initial stress map along one of the stress values of the second contact surface (2k); and a device according to claim 1 for determining the local alignment The first initial strain map and/or the second initial strain map and/or the first initial stress map and/or the second initial stress map may be considered by the evaluation component during the error. A first contact surface (1k) for bonding one of the first substrate (1) and a second substrate (20) receivable on a first platform (10) 2) a device of a second contact surface (2k) having the feature that the first alignment key positioned along the first contact surface (1k) can be detected by the position detecting member (7, 8) a first position map of (3.1 to 3.n, 50'), wherein the second alignment key positioned along the second contact surface (2k) can be detected by the position detecting member (7, 8) (4.1 a second location map to 4.n, 50), such as the device of claim 1 or 2 or the device of claim 3, for determining the value based on the value of the first location map and the value of the second location map a first alignment position of the first contact surface (1k) and a second alignment position of the second contact surface (2k) for aligning the first contact surface (1k) to the first An alignment member and an alignment member for aligning the second contact surface (2k) to the second alignment position for connecting the first substrate (1) and the second substrate (2), and A device as claimed in claim 1 or 2 or a device as claimed in claim 3. A device according to claim 4, wherein there is a detecting member for inspecting the first alignment position and/or the second alignment position during the in-situ bonding of the substrates (1, 2). The device of claim 5, wherein the inspection is performed through the transparent region of the first substrate (1) and/or the transparent region (40) of the second substrate (2). The device of claim 5 or 6, wherein there are at least two, preferably four, corresponding alignment keys (3.1 to 3.4, 4.1 to 4.4) at the time of inspection. A method for determining a local alignment error due to the first substrate (1) being opposed to the second substrate when a first substrate (1) is bonded to a second substrate (2) 2) caused by strain and/or distortion, the method has the following steps, in particular having the following sequence: detecting the strain value along the first contact surface (1k) of one of the first substrates (1) by the detecting member a first strain map and/or a first stress map along a stress value of the first contact surface (1k); and/or detecting along a second contact surface (2k) by the detecting member a second strain map of one of the strain values and/or a second stress map along one of the stress values of the second contact surface (2k); and evaluating the first strain map and/or the second strain map by the evaluation member And/or the first stress map and/or the second stress map and determining a local alignment error. The method of claim 8, wherein the first position map and/or the second alignment key (4.1 to 4.) of the first alignment key (3.1 to 3.n, 50') are considered at the time of evaluation, especially at the time of the determination. n, 50) second position map. The method of claim 8 or 9, wherein the additional step of: detecting the first initial strain pattern and/or along a strain value along a first contact surface (1k) by the initial detecting member prior to bonding One of the first initial stress patterns of the stress value of the first contact surface (1k); and/or one of the strain values along the second contact surface (2k) detected by the initial detecting member before bonding a second initial strain map and/or a second initial stress map along one of the stress values of the second contact surface (2k); and considering the first initial strain map by the evaluation member when determining the local alignment error Or the second initial strain map and/or the first initial stress map and/or the second initial stress map. The method of claim 8 or 9, wherein the following further step is: detecting the first alignment key positioned along the first contact surface (1k) by the position detecting member (7, 8) before the bonding (3.1) a first position map to one of 3.n, 50'), detecting a second alignment key positioned along the second contact surface (2k) by the position detecting member (7, 8) before bonding (4.1 a second position map to one of 4.n, 50), determining, by the computing member, a first alignment position of the first contact surface (1k) based on the value of the first position map and the value of the second position map And a second alignment position of the second contact surface (2k), the first contact surface (1k) being aligned with the first alignment position and the second contact surface (2k) by an alignment member Aligning with the second alignment position; and bonding the first substrate (1) and the second substrate (2). The method of claim 8 or 9, wherein the following further steps are performed by detecting the component, in particular through the first substrate (1) and/or the transparent region (40) of the second substrate (2), in particular The first alignment position and/or the second alignment position are inspected during bonding of the substrates (1, 2). A use of a device as claimed in claim 1 or a device as claimed in claim 4 for a reworkable or non-reworkable bonded wafer (1, 2). A use of the method of claim 8 for a reworkable or non-reworkable bonded wafer (1, 2).